| { |
| "best_metric": 0.9417759590335313, |
| "best_model_checkpoint": "model_saves/deberta-v3-large_lemon-spell_5k_2_p3/checkpoint-268", |
| "epoch": 4.0, |
| "global_step": 1072, |
| "is_hyper_param_search": false, |
| "is_local_process_zero": true, |
| "is_world_process_zero": true, |
| "log_history": [ |
| { |
| "epoch": 1.0, |
| "eval_accuracy": 0.9417759590335313, |
| "eval_loss": 0.4166739881038666, |
| "eval_runtime": 4.5145, |
| "eval_samples_per_second": 970.655, |
| "eval_steps_per_second": 7.753, |
| "step": 268 |
| }, |
| { |
| "epoch": 2.0, |
| "eval_accuracy": 0.9408032290164245, |
| "eval_loss": 0.436762273311615, |
| "eval_runtime": 4.5124, |
| "eval_samples_per_second": 971.092, |
| "eval_steps_per_second": 7.756, |
| "step": 536 |
| }, |
| { |
| "epoch": 3.0, |
| "eval_accuracy": 0.9406634689564954, |
| "eval_loss": 0.46340474486351013, |
| "eval_runtime": 4.5162, |
| "eval_samples_per_second": 970.279, |
| "eval_steps_per_second": 7.75, |
| "step": 804 |
| }, |
| { |
| "epoch": 3.73, |
| "learning_rate": 1e-05, |
| "loss": 0.2655, |
| "step": 1000 |
| }, |
| { |
| "epoch": 4.0, |
| "eval_accuracy": 0.9400988383143819, |
| "eval_loss": 0.5008934140205383, |
| "eval_runtime": 4.5059, |
| "eval_samples_per_second": 972.492, |
| "eval_steps_per_second": 7.768, |
| "step": 1072 |
| }, |
| { |
| "epoch": 4.0, |
| "step": 1072, |
| "total_flos": 1.3570052787798016e+16, |
| "train_loss": 0.2601820764256947, |
| "train_runtime": 464.6662, |
| "train_samples_per_second": 1107.376, |
| "train_steps_per_second": 8.651 |
| } |
| ], |
| "max_steps": 4020, |
| "num_train_epochs": 15, |
| "total_flos": 1.3570052787798016e+16, |
| "trial_name": null, |
| "trial_params": null |
| } |
|
|
